Fast Rule-Based Clutter Detection in Automotive Radar Data

Abstract

The unique properties of radar sensors, such as their robustness to adverse weather conditions, make them an important part of the environment perception system of autonomous vehicles. One of the first steps during the processing of radar point clouds is often the detection of clutter, i.e. erroneous points that do not correspond to real objects. Another common objective is the semantic segmentation of moving road users. These two problems are handled strictly separate from each other in literature. The employed neural networks are always focused entirely on only one of the tasks. In contrast to this, we examine ways to solve both tasks at the same time with a single jointly used model. In addition to a new augmented multi-head architecture, we also devise a method to represent a network's predictions for the two tasks with only one output value. This novel approach allows us to solve the tasks simultaneously with the same inference time as a conventional task-specific model. In an extensive evaluation, we show that our setup is highly effective and outperforms every existing network for semantic segmentation on the RadarScenes dataset [1].

Radars for Autonomous Driving: A Review of Deep Learning Methods and Challenges

“…Lidars capture limited points from the side surfaces of these vehicles when they are directly ahead. By contrast, radars yield dense uniform returns from these agents, owing to their large metallic bodies and underbody reflections [68], as discussed in Section VIII. Nevertheless, the associator in a late fusion system may routinely disregard shape features from radar model outputs due to the absence of in-context learning.…”

“…Multipath effects for autonomous vehicles can be classified into three types: double bounce, underbody reflection, and mirrored ghost detections [68]. Radar double bounces occur due to two back-and-forth reflections between an object and the radar-equipped ego vehicle, resulting in false radar observations at double the range and velocity relative to the real object.…”

Towards Deep Radar Perception for Autonomous Driving: Datasets, Methods, and Challenges

“…In the latter case, since ghost detection has similar dynamics to the real target, it is difficult to eliminate them in the traditional detection pipeline. The multi-path effect can be classified into three types [196]. The first type is the reflection between ego-vehicle and targets.…”

“…Unlike clutter, ghost objects cannot be filtered by temporal tracking because they have the same kinematic properties as real targets. Instead, they can be detected by geometric methods [196,198]. With a radar ghost dataset, it is also possible to train a neural network for ghost detection, such as PointNet-based methods [89] and PointNet++-based methods [197,199].…”

Misspecified Cramér-Rao Bound for Multipath Model in MIMO Radar

“…Therefore, automotive radars always operate in the presence of elevation multipath [24], [27][28][29]. Elevation multipath may also occur in tunnels, below bridges, or over-path road signs and constructions [30]. Horizontal multipath occurs when driving near guardrails, buildings, and adjacent vehicles [31,32].…”